Tichromatic bilinear models have been used extensively in formal approaches to color vision. A trichromatic bilinear model of visual response to reflected light can be described mathematically as a triple of 3 x 3 matrices beta 1,beta2,beta3, one for each type of photoreceptor. Each matrix maps a 3D illuminant vector to a 3D vector of reflectance descriptors. Equivalent models are related by a basis transformation X, corresponding to a choice of basis functions used to describe surface spectral properties, a transformation Y reflecting the choice of basis used to describe illuminant spectral properties and an arbitrary linear combination epsilon of photoreceptor responses.; A complete partition of trichromatic bilinear models into equivalence classes is achieved by using methods of Algebraic Geometry and Linear Algebra and a canonical representative for each such class is provided. |